You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The setup needs to be investigated. Locally I get significantly better results with Forest (average latency 2.0419 ms vs 1.1399 ms), Under heavy load (1000 concurrent requests). I'm puzzled where the reported 20 ms come from.
❯ OHA_REQUESTS=10000000 OHA_CONCURRENCY=1000 bash rpc_bench.sh
method: eth_blockNumber
params: []
payload: {"jsonrpc":"2.0","method":"eth_blockNumber","params":[],"id":1}
lotus: http://127.0.0.1:1234/rpc/v1
forest: http://127.0.0.1:2345/rpc/v1
=== sanity probe ===
lotus -> {"id":1,"jsonrpc":"2.0","result":"0x388de9"}
forest -> {"jsonrpc":"2.0","id":1,"result":"0x388de9"}
=== oha ===
--- lotus (http://127.0.0.1:1234/rpc/v1) ---
warmup (200 requests)...
Summary:
Success rate: 100.00%
Total: 20466.0296 ms
Slowest: 56.8063 ms
Fastest: 0.0243 ms
Average: 2.0419 ms
Requests/sec: 488614.5567
Total data: 429.15 MiB
Size/request: 45 B
Size/sec: 20.97 MiB
Response time histogram:
0.024 ms [1] |
5.703 ms [9588427] |■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■
11.381 ms [290064] |
17.059 ms [94021] |
22.737 ms [20312] |
28.415 ms [4672] |
34.094 ms [1355] |
39.772 ms [593] |
45.450 ms [342] |
51.128 ms [140] |
56.806 ms [73] |
Response time distribution:
10.00% in 0.2766 ms
25.00% in 0.8377 ms
50.00% in 1.6968 ms
75.00% in 2.4047 ms
90.00% in 3.7336 ms
95.00% in 5.2004 ms
99.00% in 12.3937 ms
99.90% in 21.2307 ms
99.99% in 35.3645 ms
Details (average, fastest, slowest):
DNS+dialup: 15.2847 ms, 0.2303 ms, 48.1882 ms
DNS-lookup: 0.0009 ms, 0.0003 ms, 0.0440 ms
Status code distribution:
[200] 10000000 responses
--- forest (http://127.0.0.1:2345/rpc/v1) ---
warmup (200 requests)...
Summary:
Success rate: 100.00%
Total: 11446.0108 ms
Slowest: 55.2752 ms
Fastest: 0.0309 ms
Average: 1.1399 ms
Requests/sec: 873666.8298
Total data: 419.62 MiB
Size/request: 44 B
Size/sec: 36.66 MiB
Response time histogram:
0.031 ms [1] |
5.555 ms [9989140] |■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■
11.080 ms [9884] |
16.604 ms [436] |
22.129 ms [234] |
27.653 ms [0] |
33.178 ms [58] |
38.702 ms [48] |
44.226 ms [144] |
49.751 ms [0] |
55.275 ms [55] |
Response time distribution:
10.00% in 0.3469 ms
25.00% in 0.5599 ms
50.00% in 0.9673 ms
75.00% in 1.5122 ms
90.00% in 2.1382 ms
95.00% in 2.6181 ms
99.00% in 3.7451 ms
99.90% in 5.6195 ms
99.99% in 10.9020 ms
Details (average, fastest, slowest):
DNS+dialup: 17.2446 ms, 0.2500 ms, 53.3378 ms
DNS-lookup: 0.0009 ms, 0.0004 ms, 0.0242 ms
Status code distribution:
[200] 10000000 responses
Activity
eth_*methods performance #7008LesnyRumcajs commentedon May 11, 2026
The setup needs to be investigated. Locally I get significantly better results with Forest (average latency
2.0419 msvs1.1399 ms), Under heavy load (1000 concurrent requests). I'm puzzled where the reported20 mscome from.LesnyRumcajs commentedon May 21, 2026
P95 for this method is
6.2mswith average4.8ms. Closing for now.